Design approaches prior to the discovery of PAL

The starting point for the design of the first LOV-ANTAR chimera ‘LOVA’ was the question, whether the underlying signaling mechanisms of the light-regulated histidine kinase YF1 [21,87] could be applied to effector domains of different nature. Among the potential effector domains, we chose the family of ANTAR proteins because the potential acquisition of an optogenetic tool with RNA-binding output function was an attractive side effect in addition to the mechanistic traits. ANTAR proteins naturally occur in diverse combinations with putative sensor modules, including the family of PAS sensors that represent the superfamily of our blue light-sensing LOV modules. The use of a PAS-ANTAR signaling receptor as a design template and subsequent domain replacement of its sensor module by a LOV domain was therefore an obvious consideration. However, none of the PAS-sensor-comprising ANTAR candidates had been characterized up to this time, which would have complicated the choice of an appropriate fusion site as well as the conduction of functional assays due to the lack of structural insights and known RNA target sequences. Among the formerly characterized ANTAR proteins, only AmiR featured a homodimeric conformation and a coiled-coil linker element suitable for structural superposition with the YF1 linker region. This led to the first LOVA chimera developed on the basis of sequence and structural analysis of the two parental proteins (see Section 5.1.1, Figure 11).

Benefits and limitations of bacterial reporter assays

The design of synthetic photoreceptors often requires the testing of high numbers of chimeras.

Coupling fluorescence or other easily detectable readouts to chimeric protein activity in a bacterial

reporter assay enables high or medium throughput screening, allowing the simultaneous investigation of multiple protein variants without the time-consuming process of protein purification.

Despite these advantages, the use of a bacterial reporter assay also has disadvantages: In case of the LOVA chimera, the use of the β-gal reporter system indicated that the activity level of the variants tested so far, including the initially employed positive control (AmiR resi L139 to A196/end;

hereinafter termed ‘ANTAR-AmiR’), was at the same level as the negative control. The ‘Miller unit’

was defined such that the fully induced lac-operon would amount to an activity of 1000 MUs, and the non-induced level would yield around 1 MU [152]. Upon the discovery that the induction rate of offers thus an advantage compared to fluorescent outputs, as fluorescence measurements depend largely on the experimental setup. Normalization by a control from a different system is possible, but since each system requires a specific ‘gain’ setting (see following Section, ‘Flourescence-based detection’), we long assumed in the experiments on the LOVA chimeras, for which we had observed binding activity in the EMSAs, that the detection of this activity requires a higher gain than, for example, the Dusk and Dawn system [153].

However, the quantitative translation of the enzymatic activity of the employed β-gal reporter to the output signal depends on several factors, as protein expression and functional testing occur in parallel. Since both processes are sensitive to experimental parameters, such as incubation temperature and time, light intensity, inducer concentration, the plasmid backbone(s), or their simultaneous combination, the reporter assay system is rather complex, making it difficult to distinguish between non-functional protein design and inadequate protein expression. This is a major problem for the use of a reporter assay in combination with rational design approaches, which tend to involve only limited numbers of fusion proteins into their planning and testing. In case of non-functional variants with a complete loss of activity, only little information can be derived by means of a reporter assay. Moreover, the activity curves of signaling receptors within cellular networks are often non-linear due to the cooperative behavior of many signaling proteins [154]. Therefore, small alterations of an input trigger may lead to large changes of the output signal and vice versa. The generation of an output signal therefore importantly depends on the adaption to the signaling network, so that dynamic ranges assessed with the help of a reporter screening system are only comparable for protein variants within the same experimental setup. While in vitro approaches provide well-defined systems, for which single parameters can be easily controlled, the extraction of

information from a complex cellular system, as given within a bacterial reporter assay, is more difficult. As a consequence, the determination of absolute protein activities, which are a big advantage of in vitro characterization, is usually not possible with this method. Yet, in case of successful design experiments in frame of optogenetic application contexts, the direct testing of the functional light response of engineered chimeras within the target system can be more valuable than the determination of absolute protein activities. The removal of a protein from its natural context often leads to unforeseen problems during the implementation of in vitro assays; e.g., the requirement for additional cofactors for formation of the functional (dimeric or oligomeric) conformation, or the creation of a suitable ionic environment, which often demand extensive and time-consuming optimization.

Fluorescence-based detection

Fluorescence-based reporter systems are frequently the method of choice, as they do not require any further exogenous substrates for detection owing to their intrinsic chromophore structures. This enables direct monitoring over different growth phases of the cell and a high screening throughput in combination with flow cytometry techniques. [12]. The use of the pE_β-gal reporter instead of a fluorescent reporter system (such as the I1I2_DsRed construct that demonstrated the highest dynamic range among the different promoter-DsRed constructs) is primarily due to the chronological order of the experiments. This means that the improved positive control (consisting of flAmiR) was first successful in combination with the β-gal-reporter, with which the moderate number of chimeric variants within the applied rational design approach could be easily mastered. Nevertheless, the use of the fluorescent DsRed reporter would have clear advantages for the testing of a higher number of variants, e.g. when using directed evolution- or library-based design approaches [155]. Such approaches usually permit to exctract valuable information from the reporter-based detection and selection of successful chimera variants [153], e.g. through the creation of fusion-libraries of two different protein modules that resulted in the creation of chimeras with different linker properties [88].

Nonetheless, it is important to remember that the details of the measurement device (i.e. the sensitivity of the photomultiplier tube) and the setup parameters, such as excitation and emission wavelength and gain, strongly influence detected fluorescence intensity [154]. Fluorescence measurements are particularly susceptible to errors in ranges of weak gene expression, where autofluorescence often causes a high associated error [154]. Autofluorescence derives both from biological structures, as well as many non-biological materials, such as organic plastic polymers.

Therefore, at low expression levels, autofluorescence variations often cause high associated errors,

as experienced during the initial attempts to construct the pE_DsRed Reporter System with the LOVA chimeras and the erroneous positive control.

Domain replacement guided by structural superposition

The initial LOVA chimera exhibited good expression in E. coli. The qualitatively conserved spectral and photochemical properties (see Figure 13 in Section 5.1.1) indicate that the structure of the LOV module within the LOVA chimera has been preserved. Within the EMSA, the LOVA receptor demonstrated a binding affinity (500 – 600 nM) to the ami-lead transcript comparable to the initial positive control, i.e. the isolated ANTAR domain of AmiR, but no apparent difference under dark and light conditions. However, the β-gal reporter assay revealed that the activity level of the LOVA chimera, as well as the ANTAR-AmiR positive control employed so far, was at the same level as the negative control. The positive control used up to this point only consisted of the C-terminal ANTAR module of AmiR (resi L139 to A196/end). Since protein domains are commonly considered as an autonomous unit of organization that can fold independently into a stable structure and exist and function independently of the rest of the polypeptide chain [156], we assumed that this would be sufficient to maintain the RNA-binding function of the AmiR receptor and thus represent an appropriate positive control. Moreover, the via EMSA determined binding affinity of ANTAR-AmiR was in a similar range as for EutV, another recently characterized representative of the ANTAR family from E. faecalis [100]. Since no further comparable experimental data on RNA binding activity was available for AmiR, and the binding affinity of the EutV protein was in the same range as within our EMSA tests, we initially assumed that the binding affinity of ANTAR- AmiR corresponded to a binding level of the active receptor state. In addition, ligand receptor associations are usually concentration-dependent equilibrium processes. Since the physiological concentrations, under which the receptor function is optimally expressed, are often unknown, the determination of the dissociation constant Kd unfortunately does not offer an absolute measure such as ‘binding’ or ‘non-binding’ in order to evaluate the functionality of a receptor [150].

The strongly reduced or terminated reporter activity using the AmiR-ANTAR construct, compared to the full length protein, suggests that the truncation of the ANTAR domain leads to a reduction in RNA binding activity. This is in line with the report of Ramesh et al. [100] on their experiences with the ANTAR signaling receptor EutV. In this study, the RNA binding activities of three different EutV variants were examined: the full-length EutV protein (flEutV) which comprised the N-terminal receiver domain, a coiled-coil region and the C-terminal ANTAR domain; a second construct that included the coiled-coil region (ccANTAR-EutV) and the ANTAR domain; and, just the C-terminal ANTAR domain (ANTAR-EutV) (see Figure 41). The data suggested that the target RNA is bound with

decreasing affinity by ccEutV, flEutV and EutV, respectively. Compared to ANTAR-EutV, the binding affinity of ccANTAR-EutV to the RNA target substrate determined via EMSA binding assays was about 100-fold higher. However, the RNA binding activity of ccANTAR-EutV was in a similar range as for the isolated AmiR ANTAR domain within our EMSA tests (KdccANTAR-EutV

≈ 0.7 µM).

Whereas flAmiR shows activities in the fully induced range in the β-gal reporter assay, the binding activity of the flEutV is significantly lower (KdflEutV

≈ 10 µM) than for ccANTAR-EutV within the EMSA experiments of Ramesh et al [100]. This fact might be explained by the different activation mechanisms of the two ANTAR proteins: the activation of EutV occurs by means of signal-induced phosphorylation of the N-terminal response receiver (RR) domain, which stimulates dimerization of the EutV monomers thereby enabling the association with its target RNA. The signal transfer is mediated by the corresponding sensor histidine kinase EutW that undergoes autophosphorylation in response to ethanolamine and consequently acts as positive regulator. The activation of AmiR, on the other hand, is carried out with the aid of the negative regulator protein AmiC [102]. The induction of AmiC by small-chain amides leads to release of AmiR allowing its association with its target RNA sequence. Consequently, AmiR displays a constitutive activity in absence of AmiC, whereas in case of EutV, the RR domain seems to attenuate the RNA binding activity of the ccANTAR in its unphosphorylated state in absence of EutW. However, a clear explanation for the reduced binding affinity of ANTAR-AmiR compared to the full-length receptor is not given by previous studies.

One possibility is that the RR-part AmiRs is required for the production of the functional dimer unit, but the RR might also (somehow) be required for the antitermination observed in our assay.

Figure 41: Different activation mechanisms of EutV and AmiR. Scheme for illustration of the conclusions drawn in the previous Section. Even in absence of its positive regulator AmiC, AmiR displays a constitutive RNA binding activity C, whereas in case of EutV, the RR domain of EutV seems to attenuate the RNA binding activity of the ccANTAR module. (a) Overview of truncated contructs from the study of Ramesh et al. [95] together with the via EMSA assessed affinity constants (Kd) to their target sites. (b) Overview of the truncated AmiR contructs tested within this study. RR – common response regulator receiver domain; RR* - pseudo RR, i.e., residues essential for phosphoryl acceptance lacking.

Effect of linker-length variation

As an insufficient length of the coiled-coil linker might prevent the formation of the functional dimer, we decided to assess if an elongation of the N-terminal coiled-coil region of the AmiR ANTAR domain would improve its RNA-binding affinity. For that purpose, additional Miller assays with extended versions of the initial positive control (ANTAR-AmiR), comprising the ANTAR core module and increasing parts of the coiled-coil (see Figure 19 in Section 5.1.1), were performed to assess if these changes would result in higher reporter activity. However, even the longest of these variants (‘+31-ccANTAR-AmiR’; see Table A3 in Section 8.2) that included the full extent of the AmiR coiled-coil region did not result in an increase of the output signal. We hence assumed that further shifting of the fusion site towards an elongated AmiR linker would most likely not improve the binding characteristics. The potential reasons for the persistent inactivity of the LOVA chimeras are diverse:

first of all, the light-induced structural signal elicited by the LOV domains might not be compatible with the activation mechanisms of the AmiR ANTAR effector. Another potential reason might be an unfavorable choice of the selected fusion site. Whereas PAS/LOV signaling receptors feature the highly conserved DIT motif that marks the C-terminus of the PAS/LOV core domain, the AmiR coiled-coil linker region that connects the N-terminal (pseudo-) RR domain with the C-terminal ANTAR domain does not feature such a clear boundary. Moreover, the C-terminus of the coiled-coil α-helix features interactions with the three-helical bundle that defines the ANTAR core domain. To maintain these interactions, the fusion site had to be selected further down the coiled-coil linker region. For YF1, it was shown that some of the N-terminal residues of the Jα coiled-coil linker element are essential for the transmission of the activating light stimulus, therefore the fusion site had to be set inside the α-helical motif of the linker region for both sensor and effector. This process is error-prone and can easily lead to irregularities within the linker helix, which would subsequently impair the transduction of the intramolecular signal. Further investigations would be required to clarify which of the discussed explanations is actually valid. Considering that the activity of AmiR within its natural context is controlled by the positive regulator AmiC, differences between the signaling mechanisms are conceivable.

Besides, the linker length analysis for the family of PAS-ANTAR signaling receptors did not reveal any conserved pattern (see Figure 18 in Section 5.1.1), as found for the family for PAS-HisKA. The linker regions of the PAS-HisKA family display a significant heptad periodicity regarding their lengths, which is a result of the preservation of the hydropathy pattern characteristic for coiled-coil elements [35,157]. For YF1, the maintenance of this heptad-periodicity was found to be crucial to establish light-dependent kinase activity. In contrast, in case of the LOVA chimeras, the alteration of the linker length between sensor and effector module, which led to light-regulated constructs in other photoreceptor-engineering approaches, failed to improve the RNA-binding function and

light-sensitivity of the chimeras. The linker variants of the tested LOVA chimeras encompass linker lengths variations of 14 residues, which corresponds to the insertion of two helical turns between the - 4 and +10 –LOVA contructs (see Table A3 in Section 8.2). For the shortened linker variants of the initial LOVA chimera (-1 to -4 – LOVA; see Table A3), the remaining residues of the AmiR linker sequence C-terminal to the start of the ANTAR core domain were consecutively reduced until the beginning of the conserved three-helical bundle. The alternative reduction of N-terminal residues that belong to linker residues of YF1 would have eliminated essential residues, for which single point mutations (e.g.

for the residues D21 or V27) were found to have severe effects on signal transduction within YF1 [21,155]. For that reason, a further reduction of the linker length seemed not beneficial. On the other hand, the successive elongation of the linker by 10 residues did not result in any change in activity or light-sensitivity, so that we decided to cease the efforts of designing a light-regulated ANTAR chimera based on the superposition of YF1 and AmiR at this point. Still, further systematic analysis would be required to assess if the lack of activity of the tested chimeras is due to mechanistic differences or to an unfavorable design. One notable further option could be to shift the original fusion site more towards the C-terminus of the linker sequence without moving the overlay grid further, e.g. from E138 to L143. This would be particularly interesting, since Gleichmann et al. [155] observed that point mutations of the YF1 linker residues E142 and L143 led to constitutive kinase activity irrespective of the light conditions. However, unlike the constitutively active YF1 mutants, the so far tested LOVA chimeras have shown no signs of activity within the reporter assay up to now and, and only low affinity to their target RNA in the case of the initial LOVA chimera tested via EMSA. For this reason, we decided to adopt a new strategy instead.

Use of associating LOV modules

As outlined in Section 3.3.2, the use of associating photoreceptors has been proven quite successful for optogenetic engineering attempts in the past, even for effector types that were not regulated by association- or dissociation-based processes before. For that reason, we decided to employ another design approach for building a light-responsive ANTAR protein, by using a LOV domain type for which the transmission of the light signal was shown to occur upon association-based mechanisms.

With (i) the LOV domain from N. crassa Vivid (‘VVD-LOV’), and (ii) the LOV domain from the N.

gaditana Aureochrome (‘NgAur-LOV’), two different LOV modules were fused at the N-terminus of two different fusion sites of the ccANTAR module from P. aeruginosa AmiR via a short and flexible linker, resulting in four distinct LOV-AmiR chimeras. As additional control, the VVD LOV domain was attached to the N-terminus of the full-length AmiR protein via the same linker (‘VVD-flAmiR’) in order to investigate if this would interfere with the functionality of the ANTAR effector. Only this last

construct resulted in a detectable signal, even though the activity level was sharply reduced compared to the positive control consisting of flAmiR alone. The signal of the remaining associating LOV-AmiR variants was at the level of the negative control, regardless of the light conditions. Only little information can be drawn from these results. The lack of a signal within the reporter assay can have several reasons, as detailed in the previous Section (‘Benefits and limitations of a reporter assay’), e.g., problems with folding of the protein structure, consequently leading to a reduced yield of protein expression. This is not unlikely, especially for the S90 variants whose fusion point lies within the RR unit. However, other scenarios are conceivable, e.g. the AmiR-RR might be required for

construct resulted in a detectable signal, even though the activity level was sharply reduced compared to the positive control consisting of flAmiR alone. The signal of the remaining associating LOV-AmiR variants was at the level of the negative control, regardless of the light conditions. Only little information can be drawn from these results. The lack of a signal within the reporter assay can have several reasons, as detailed in the previous Section (‘Benefits and limitations of a reporter assay’), e.g., problems with folding of the protein structure, consequently leading to a reduced yield of protein expression. This is not unlikely, especially for the S90 variants whose fusion point lies within the RR unit. However, other scenarios are conceivable, e.g. the AmiR-RR might be required for